What a fun week it has been this week with three issues on quantum computing technology. I hope you have enjoyed Quantum Week at The Bleeding Edge.

I’m excited to dive even deeper into the quantum opportunity next Tuesday, as well as to discuss the top three quantum companies I’ve got my eye on.

We’ve hit an inflection point in quantum computing in the last 12 months. As a result, the industry is now on a path of accelerated growth with quantum computers already performing feats impossible for classical supercomputers.

Some of the biggest names I’ve been watching and writing about over the years – companies like Rigetti (RGTI) and IonQ (IONQ) – have had a stellar year. No doubt, they’ll do amazing things from here.

But they’re not my favorite companies to buy right now. And we’ve only scratched the surface of the opportunity that quantum computing and the race towards fault-tolerant quantum computing systems present…

You can go here to sign up to join me next week on October 21, at 8 p.m. ET, to hear more about it. We really do have so much to look forward to. In quantum computing, certainly… but quantum computing wasn’t the only interesting story this week.

If things couldn’t get any bigger in AI, OpenAI and Broadcom announced a massive deal to deploy 10 gigawatts (GW) of custom artificial intelligence chips over the next four years, which added about $150 billion to Broadcom’s market value.

Citi came out with a bold announcement that it will begin custody services for cryptocurrencies in 2026, an unexpected move from the massive financial services company that has been way behind on the adoption of and support for digital assets.

Even crazier was the coalition of NVIDIA (NVDA), Microsoft (MSFT), xAI, and BlackRock stepping up to acquire data center firm Aligned Data Centers for a whopping $40 billion. If you can’t build them fast enough, just buy them…

And SpaceX had a remarkably successful 11th test flight of the Starship into space. It was the last test flight of the current iteration of the Starship. Future flights will incorporate new design aspects into the Starship in preparation for what will quickly become SpaceX’s regular commercial Starship flights to orbit every month in 2026.

It’s incredible how exciting 2025 has been, and it’s even harder to believe that 2026 will be far more incredible.

Have a wonderful weekend,

Jeff

Quantum in Tandem?

Hello Jeff,

As a lifetime member of your service, I thoroughly enjoy reading all your reports.
I have a question, which might be a stupid one, but I need to ask you:

Could we imagine some scenario where a quantum computer and a traditional GPU server work in tandem? In biology, for instance.

The quantum computer does the rough calculations to eliminate 98% of the cases, and then the traditional server analyzes the remaining 2% of the cases…

Benefits would include processing time, energy cost savings… What are your thoughts about it? Thank you.

– Paul B.

Hi Paul,

That’s definitely not a stupid question at all. It’s a smart one. There is a lot of nuance in terms of how best to use different forms of computing systems.

A very simple example is the difference between the computing systems that are used for training frontier AI models versus those that are designed for inference (i.e., the running of said AI models).

Today, data centers comprised of hundreds of thousands of NVIDIA (NVDA) or AMD (AMD) GPUs combined with Intel (INTC) or AMD CPUs are most widely utilized for training frontier AI models.

However, it is far more efficient to run those models on AMD GPUs, which are optimized for inference, or on other up-and-coming application-specific semiconductors from companies like Cerebras, Groq, SambaNova, or d-Matrix.

The same concept is true when we think about how best to utilize a quantum computing system versus an AI data center primarily made up of NVIDIA or AMD GPUs. We explored this concept a bit in yesterday’s Bleeding Edge – Hybrid Quantum.

Coherence times for quantum computing systems are still quite short. They can’t run for hours or days on end. So, the best utilization of a quantum computing system is to give it the most complex part of the puzzle to solve – the part that is impossible for an AI data center to solve in any reasonable period of time.

In yesterday’s example, the hardest part was solving for the ground-state energies of a molecule. Once a quantum computer has cracked the problem, then the remainder of the computational tasks can be offloaded to a GPU-based supercomputer to complete the overall task.

When it comes to quantum computers versus GPU-based data centers, it’s more about using the best system to complete the task.

But to your point, when it comes to training a frontier model versus inference, application-specific semiconductors are highly desirable because they have such significant energy cost savings per unit of compute (i.e., they are far more efficient at their task). And energy efficiency when running AI applications is the most important factor impacting gross margins and end pricing to end users.

A good framework to think about this interrelationship between the two different kinds of computing systems is that quantum computers won’t replace GPU-based AI data centers. They will enable us to solve complex problems that otherwise wouldn’t be possible.

They are accelerants to the exponential growth that we are witnessing right now. AI and quantum computing are the catalysts that are pulling us into a phase of hyper-acceleration in terms of technological advancements and economic growth.

What’s Going on With the Solar Industry?

Hi Jeff,

Thank you for all your insights, they have truly helped me look past all the noise of the news and really understand where we are going and how we can get there.

I was hoping you could provide your insight on the recent cancellation of the Esmeralda 7 project, and more specifically, what the purpose of this was and how this impacts the solar sector moving forward.

Additionally, I was hoping you could touch on solid oxide fuel cells and how they might play a role in the upcoming energy demand boom.

Thanks again for all your work.

– Kirk R.

Hey Kirk,

This is an interesting topic for a variety of reasons.

For everyone’s benefit, the Esmeralda 7 solar project is a cluster of seven large-scale solar installations, plus associated energy storage systems to be built on 62,000 acres of land belonging to the Federal Bureau of Land Management (BLM) in Esmeralda County, Nevada.

If the entire project is completed, it could have produced up to 6.2 GW of electricity under good conditions, and 5.2 GW of battery storage. This would represent the largest solar installation in the U.S.

It appears that there has definitely been some confusion over what has happened with the Esmeralda 7 project. The BLM had not yet approved the project to move forward.

The media has tried to position these latest developments as the project being canceled, but the U.S. Department of the Interior has denied that this is the case.

What was canceled was the permitting of the project from a broad, single environmental review. What is being permitted is for each of the seven individual projects to conduct their own environmental impact reviews and for the developers to pursue the individual permits for each of the seven projects.

In short, they can still move forward.

It is worth mentioning that there was quite a bit of controversy around the Esmeralda 7 project from both the residents of Esmeralda County and some conservation groups that were concerned about the impact on wildlife habitats, among other things. Given the scale, 62,000 acres, there is some good logic to permitting the sites individually.

Deserts are one of the places where large-scale solar can make a lot of sense. A project like this, ideally, could be an opportunity for the U.S. to scale up U.S. solar panel manufacturing to support a project like this.

With scale comes lower costs, which would benefit the entire U.S. solar industry. So, if the economics make sense, hopefully this series of seven projects moves forward with the backing of the developers behind this project, like NextEra Energy, Leeward Renewable Energy, Arevia Power, and Invenergy.

What I don’t want to see is another corruption-fueled, taxpayer-funded mess like Solyndra.

Solyndra was a solar company that shut down in 2011 after it repeatedly misled federal officials and fudged the firm’s financials to secure a major half-billion-dollar-plus government loan at taxpayers’ expense.

Ironically, one of the best things that could happen for the solar industry would be the decline of interest rates. When the cost of capital declines, large-scale infrastructure projects are easier to fund. The economics improve.

As for the solid oxide fuel cells, I actually touched on this topic last month in The Bleeding Edge – CLARITY Is Coming. It was another AMA issue, and the question was specifically about Bloom Energy’s (BE) fuel cells, but the first response is very relevant to your question.

It’s important to remember that these fuel cells are fueled by fuels like natural gas or hydrogen, which are either fossil fuels or made from fossil fuels. These fuel cells can be part of the solution for providing electricity to meet industrial energy demands.

One key point that I made in the previous AMA was that, because there is such a shortage of natural gas turbines, fuel cell companies have been able to fill some of that gap in supply with their own technology, despite the extremely high costs of using solid oxide fuel cells.

The big disadvantage to fuel cell technology is the high costs, so this current resource-constrained market environment has actually created a market for these companies. It would be a different story if natural gas turbines were widely available and didn’t have multi-year backorders.

There will, however, still be a future market for these kinds of fuel cells, as they give the veneer of being “green.”

Power is generated electromechanically without any combustion, which results in lower overall emissions compared to a combustion process. It’s still fossil fuels, but a lower-emissions option, albeit at a much higher expense.

Nowhere near as green as many would have you believe, but ultimately not a terrible option to fill in some gaps in the power supply as energy demand continues to climb…

Inversion Space: Retrievers of the Lost Arc

Love reading your articles about “far-out” ideas that somehow eventually become not only possible, but widely used.

Your article about Inversion Space’s Arc delivery system is another fine example. You indicated that the Arc is reusable, but did not explain how.

Someone or something will have to retrieve it. If it lands somewhere extremely remote, retrieval would be very expensive. Thanks.

– Jeff H.

How do they retrieve the Arc from the remote location for reuse?

– Arthur R.

Hello Jeff and Arthur,

You’ve raised a very practical question about how Inversion Space’s Arc spacecraft would be used in practice. I’m sorry I didn’t address this when I wrote about it in The Bleeding Edge – 1-Hour Delivery, Anywhere on Earth.

As a reminder, the Arc spacecraft is about four feet wide and eight feet long… and small enough to sit on top of a conference room table. It has been designed to transport up to 500 pounds (225 kg) of payload/packages anywhere on the planet.

Arc’s Dimensions | Source: Inversion Space

As you both stated, Arc is designed for reusability, but what happens after it lands at its desired location? How do they recover it for reuse?

Well, there is no magic wand or secret here. The spacecraft does have to be physically recovered and returned to Inversion Space to be turned around for the next flight. And there are three factors that Inversion took into consideration when designing Arc for this purpose:

Arc uses non-toxic propellants, which means that it is safe to retrieve the Arc immediately after landing, and safe to work with the Arc to get it loaded onto some form of transportation for recovery.
Arc is designed to be small enough to be loaded onto a truck, into a helicopter bay, rail car, a cargo plane, or even a boat. While there are logistics involved, anywhere there is a forward operating team, there are certainly modes of transport in and out of that location.
Arc can autonomously land at any designated target. That also means that the landing zone can be chosen based on having some form of recovery route in mind, whether it be air, land, or sea.

The other key point is that for the applications for which Inversion’s spacecraft will be used, it is the quick, 1-hour delivery that is important. That’s what’s time sensitive. The recovery of each Arc is less time-sensitive, which gives a lot more flexibility for lower cost, less time-sensitive recovery.

And all things considered, the cost to load something onto a truck or helicopter, return it to a military base or central location, and then return to Inversion will be a small fraction of the combined build, launch, and delivery costs associated with this kind of technology.

It may be far out there, but it’s happening now. I’m excited to see this kind of technology in operation.

Jeff

Want more stories like this one?

The Bleeding Edge is the only free newsletter that delivers daily insights and information from the high-tech world as well as topics and trends relevant to investments.